Backtest Before You Bet Real Capital

[FTtrader]

Before committing real money to any new scalping idea, put in the work to backtest it across a genuinely meaningful sample — at least 100 historical setups, and ideally more, across a range of different market conditions rather than just one particularly favorable stretch.

A strategy that "feels right" after five live trades has told you essentially nothing statistically meaningful. Five trades is well within the range where pure random variance can produce a misleadingly strong (or misleadingly weak) result, regardless of whether the underlying strategy has any genuine edge at all.

Proper backtesting means going back through historical charts, applying your specific, written entry and exit criteria mechanically and honestly — not cherry-picking the setups that would have worked while conveniently skipping the ones that wouldn't have — and tracking the results with the same rigor you'd apply to live trades: win rate, average reward-to-risk, maximum consecutive losses, and overall expectancy.

This process takes real time and effort, which is exactly why so many traders skip it and jump straight to live trading with real capital instead. But the traders who consistently survive and improve over the long run are, almost without exception, the ones who did this unglamorous groundwork first, rather than discovering a strategy's flaws the expensive way, one live losing trade at a time.

Hello traders, fairman,

i think some strategies you can not backtest, because you will not have that News at that time plus fundamental bias.

[LondonScalper]

Before committing real money to any new scalping idea, put in the work to backtest it across a genuinely meaningful sample — at least 100 historical setups, and ideally more, across a range of different market conditions.

Hundred-plus across regimes is the right bar. “Feels good on last week” is not research.

My process: define rules in writing, sample through trend and range weeks, include costs, then forward-test small before full size. Indicator soup that only works after optimisation is a curve-fit wearing a lab coat — your later exchange on pure TA combinations matches what most of us find when we are honest.

Desk rule: a new idea stays in a named sandbox account or half-size tag until the live sample agrees with the historical sketch. AI can speed coding; it does not excuse skipping the ugly weeks in the test.

Backtest to learn failure modes, not to mint certainty. If the method needs perfect fills and perfect regimes, it is not ready for the London open.

[PropScalpDesk]

Before committing real money to any new scalping idea, backtest it across a meaningful sample — at least 100 historical setups, preferably more.

Sample size is respect for variance. From this desk a new idea gets markup and a cost-adjusted sample before live micro size. One pretty week on demo is not a green light; it is a suggestion to keep measuring.

Practical rule: I will not promote an idea to funded risk until personal micro risk shows the same behaviour under my spreads. Prop challenges are for executing a known playbook, not discovering one.

Also force the ugly periods into the sample — news weeks, dead Asia handoffs, wide Monday opens. Edge that only works on calm Tuesdays is not an edge.

I also forward-test with a written invalidation of the idea itself: if live micro diverges from backtest after N tickets, the idea is demoted. No “it will warm up.”

What is your minimum sample before first live cent/micro lot — 100, or do you require more on gold?